Cheap simple class D amp circuit to build.

Status
Not open for further replies.
Jump to Latest
#141
My drivers stay cool at all times. The main heatsink is a bit above warm, but i don't know what it is as i've put all devices on it (regs and mosfets). The linears were getting very hot on the stripboard. The temperature is pretty much the same (55ish C) regardless if it's loaded or idle so i guess it's the regs.

I remembered i had a 4.3 ohm wirewound dummy load which is good for a few hundred watts short-term and about 100W long-term. Yay, no more waking everybody up with the noise. This also gave me the opportunity to test clipping at all frequencies. Above 800Hz or so the output clips cleanly. As the frequency is lowered a strange waveshaping phenomenon happens which is the source of the ugly distortion i've been hearing. At very low frequency, when clipping threshold is reached, the waveform cuts out and goes to ground. I'll be looking for possible causes now.
 
#142
It now comes to mind that i never tried a simulation at low frequencies. So i gave it a go, and bam, i get the same type of distortion. Which confirms that it's a problem with the concept not the physical circuit.

A while ago someone suggested to me that the amp "stops oscillating" when it's clipped. This is true in a sense as when this thing is driven into clipping a very long pulse is created for the duration of the clipping. But this has nothing to do with the distortion i am seeing. However - very long pulse = DC. Gate driver cannot pass DC so mosfet discharges and turns off because of 10k resistor. Here we have our problem.

A number of pages back someone suggested that the gate driver caps should be changed to 1uF. This is a possible solution indeed and i'll be trying it now. Edit: But why do we have to discharge the mosfets in the first place? Well... we don't. So i cut the 10k resistors et voila. Clean clipping. Now, since the mosfet cannot discharge, in the event that the thing does stop oscillating there is a major risk of the output sticking to one of the rails and frying the speakers. But, we do not need to go to such extremes. We just need to change out the 10k resistor for a much higher value. Time for some math, i'll be right back.
 
Last edited:
#143
I pulled out the meter and measured in-circuit the capacitance between the gate and source of my mosfets. I got about 4.9 nF on both. Now, plugging this into a time constant calculator with 10k results in 49 microseconds. Which means, that when the duration of any single pulse exceeds 49 microseconds, the mosfet will turn off. 49 microseconds resolves to 20.5kHz. When clipping a low frequency wave it's kinda obvious that the output will stay railed for way, way more than that. It would've been all fine and dandy if this were operating a limiter, but what it does is introduce uglier distortion than clipping itself. It does limit the output power alright, but does so in a very nasty way.

So change the 10k resistors to 470k or 1Meg. This will allow you to rail it as much as you like and still turn off the mosfets in an useful amount of time (2.3ms/4.9ms) if the driver section fails.
 
#146
luka said:
or you add circuit, that will lower gain when clipping, this way you'll never hard clip for long time
Click to expand...

That's in the works. But still, no limiter can ensure an attack time of under 50 microseconds unless you delay the audio. So the 10k resistors would still need to be changed for higher values.

Pafi said:
With additional 2 BJTs and 2 resistors DC coupling of gate signal can be done.
Click to expand...

It's not needed at this moment but i'll try it later on and see if it improves anything.
 
#149
Th3 uN1Qu3 i`ll try to change 10k to 470k tonite and will test it.

Pafi, (With additional 2 BJTs and 2 resistors DC coupling of gate signal can be done), can you tell me the details about this one? how to connect them, circuit details? i`m willing to test it.

regards,
Azmi.
 
#150
Aucosticraft said:
Can you pls post simulation folder. so that we too can get to visualize it and see the parameters.

EDIT : What is your supply voltage? and peak input voltage?
Click to expand...

See attachment. Peak input voltage isn't relevant as i have put in a trimpot for the gain (instead of R6 22k i put in a 50k trimpot) and so the gain can be adjusted to suit various sources. Supply voltage is +/-30v not regulated, from a mains transformer.

Simulation is based on the one posted in this thread, only that i've made the power supply more real-world, changed mosfets to models already available in LTSpice, modified output filter for less distortion (it's the same as i have in the real thing, well, at least the capacitor is, as i know nothing about the inductor i am using except that it came from the 3.3v output of an ATX power supply, it's double layer on a ferrite bar, and it's the only one i found in my junk box that doesn't heat up like crazy, i have three of these, rest are only -26 and -52 material), changed 8v2 zeners to 12v (since the driver supply is +/-12v anyway), and increased input level to the standard pro audio +4dBu. Not that input level would matter too much tho, it's just for my own convenience.

Run the sim, watch the super ugly distortion, then change R5/R6 to 1Meg and run again.

Pafi said:
Think it over, what happens if supply rail changes during clipping! Do you still think you don't need it?
Click to expand...

Well, the final version will be fed from an unregulated SMPS as it's the best way to maximize power at 8 ohms and limit the current at low impedances. Plus, it requires less math. 😛 If i have no issues with the amp running on that supply, then i don't need it.

norazmi said:
Th3 uN1Qu3 i`ll try to change 10k to 470k tonite and will test it.
Click to expand...

Awaiting results. On mine this worked great. It may still do it at very low frequencies, so use 1Meg or larger if it still happens. But i think you'll be past the mechanical limits of your speaker before the amp clips at 20Hz. 😀
 

Attachments

Last edited:
#153
albert_emule said:
Someone here already construio the Feedback atavés of low raisin filter to an invéz of catching the sign of the inductor of exit?


abaixe o arquivo anexado. Ele tem a simulação.
Não esteja em uma pressa, ele é atrasado para simular
Click to expand...

correcting:

Someone here already interconnected Feedback with pass low filter, disconnecting of the sign of the inductor output?
 
#154
Not yet, idk the others, i`ve tried with smps no noise so far, and its powerfull, rail voltage +-45 vdc, mosfet remains cool, measured for opamp/ comparator and 2 bip-trans remains cool too, supply voltage is stable with 8.90 vdc using esp model regulator tip41 and 42 feed with 9.1v zener and 2.2k 3w res.

i would like to test with 4 ohm later, any suggestion for the lc filter value and caps?

regards.
azmi
 
#156
Hi Th3 uN1Qu3,

I have some result with this baby, i can low down thd up to 0.14%, base on your schematic with LTspice, also i do record with camera when it push maximum volume with my N82 HP, i`ll post here later...

Circuit: * D:\DX-Amp\d-amp projects\cheapd-sim-clipping-lf\18_cheap-d-v1b-530-393.asc

Direct Newton iteration for .op point succeeded.
Fourier components of V(out)
DC component:0.0652971

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+02 7.936e+00 1.000e+00 -0.11° 0.00°
2 2.000e+02 4.522e-03 5.699e-04 -89.28° -89.18°
3 3.000e+02 1.166e-02 1.470e-03 -179.89° -179.79°
4 4.000e+02 1.947e-03 2.453e-04 -89.26° -89.16°
5 5.000e+02 5.661e-03 7.133e-04 179.67° 179.77°
6 6.000e+02 1.566e-03 1.973e-04 -91.33° -91.22°
7 7.000e+02 3.661e-03 4.614e-04 -179.81° -179.71°
8 8.000e+02 1.261e-03 1.589e-04 -89.91° -89.80°
9 9.000e+02 2.507e-03 3.159e-04 -179.58° -179.47°
Total Harmonic Distortion: 0.144231%



Date: Tue Mar 01 20:27:18 2011
Total elapsed time: 462.843 seconds.

tnom = 27
temp = 27
method = modified trap
totiter = 6435906
traniter = 6435872
tranpoints = 1437994
accept = 1016197
rejected = 421797
matrix size = 72
fillins = 135
solver = Normal
Matrix Compiler1: 11.4 KB object code size 11.9/9.0/[4.9]
Matrix Compiler2: 8.22 KB object code size 5.8/6.9/[2.9]

regards,
Azmi
 
#160
LOL, you`re right Luka 240wrms 8 ohm speaker, anyway, this is my last result, i need to rest with this circuit, hate to wait the LTspice simu, its takes time to finish it. So this will be my last result and wanna put them on case right now.

Circuit: * D:\DX-Amp\d-amp projects\cheapd-sim-clipping-lf\18_cheap-d-v1b-530-393.asc

Direct Newton iteration for .op point succeeded.
Fourier components of V(out)
DC component:0.0447465

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+02 1.103e+01 1.000e+00 -0.07° 0.00°
2 2.000e+02 5.400e-03 4.896e-04 -89.35° -89.28°
3 3.000e+02 5.880e-03 5.331e-04 179.76° 179.83°
4 4.000e+02 1.111e-03 1.007e-04 -91.12° -91.05°
5 5.000e+02 3.751e-03 3.401e-04 -178.95° -178.88°
6 6.000e+02 6.607e-04 5.990e-05 -90.93° -90.86°
7 7.000e+02 2.570e-03 2.329e-04 -178.33° -178.26°
8 8.000e+02 5.011e-04 4.543e-05 -87.35° -87.28°
9 9.000e+02 1.872e-03 1.697e-04 -178.10° -178.03°
Total Harmonic Distortion: 0.085930%



Date: Tue Mar 01 23:37:52 2011
Total elapsed time: 532.640 seconds.

tnom = 27
temp = 27
method = modified trap
totiter = 6823424
traniter = 6823390
tranpoints = 1531624
accept = 1081431
rejected = 450193
matrix size = 72
fillins = 135
solver = Normal
Matrix Compiler1: 11.4 KB object code size 11.9/8.8/[4.3]
Matrix Compiler2: 8.22 KB object code size 5.6/6.9/[3.1]
 
Status
Not open for further replies.
Top Bottom